20 Jul, 7 tweets, 3 min read
I've seen a lot today about how effective the vaccines are; mistakes aside, lots of folks seem to be mixing up which denominators matter - good thing statisticians *LOVE* denominators 🥰

1/6
If you see something like x% of the sick/hospitalized/deceased were vaccinated, the *better* the vaccine uptake the *scarier* this number will seem! It is using the wrong denominator. For example, here is a scenario with 90% vaccination, 4 people got sick: 2 vaxed 2 unvaxed

2/6
In this scenario, 50% of the sick were vaccinated, but this is the wrong metric to look at! It is using the wrong denominator. It doesn't take into account that *90%* of the population is vaccinated (yay!)

3/6
What you need to do is look at the rates among the vaccinated and unvaccinated separately, and then compare them. Here 11% of the vaccinated got sick, 100% of the unvaccinated got sick.

4/6
We calculate vaccine efficacy as

(risk among unvaccinated - risk among vaccinated) / risk among unvaccinated

so in this case, 89% (yay!)

5/6
So in sum, denominators matter! When scrolling past headlines, be sure to think about what denominators are in play

6/6
Denominators matter, in blogpost form!

livefreeordichotomize.com/2021/07/21/den…

• • •

Missing some Tweet in this thread? You can try to force a refresh

This Thread may be Removed Anytime!

Twitter may remove this content at anytime! Save it as PDF for later use!

More from @LucyStats

11 Dec 20
📅🎄 We are officially 14 days, or one #COVID19 quarantine, from Christmas

Since folks are likely considering quarantining before gathering this season @Ymax, @JustinLessler & I put together a quick explainer on why 14 days is recommended 👇

lucystats.medium.com/only-one-more-…
Where does 14 days come from?

👩‍🔬 Scientists have been collecting data on the number of days from when someone is infected until they show symptoms. They can then plot this data and look at the shape to gain an understanding of the incubation period.
There have been lots of studies that do this, and all of them show a pretty similar shape, with a peak at 4–7 days and a tail that extends out at least 2 weeks. Combining all of the data from all of these studies gives an even better estimate.
20 Jul 20

Still need to work out some production-things but I was able to make this for <\$100 with:

✅ plexiglass
✅ LED strip lights
✅ 4 bookshelf L-brackets
✅ 2 C-clamps

mostly followed these instructions:
Supplies:

◻️ 1/4 x 24 x 36 plexiglass amazon.com/gp/product/B00…
💡 16ft of LED strip lights: amazon.com/MINGER-Changin…
🗜 2 one inch c clamps amazon.com/TEKTON-Malleab…
📖 4 bookshelf brackets amazon.com/gp/aw/d/B07RHL…
🖊 neon markers amazon.com/gp/aw/d/B008DQ…

I made this in 4️⃣ steps 👇
Step 1️⃣

💡 stick 3ft of the LED strip lights on a table
☝️ your plexiglass will sit on top of this
22 May 20
A paper recently came out in NEJM on COVID-19 patients using remdesivir:

📰 nejm.org/doi/full/10.10…

Bonovas & @piov1984 wrote a great critique pointing out that the analysis had a flaw!

nejm.org/doi/full/10.10…

This thread is on *competing risks* & why it is so important! 1/13
First let's talk about the *question* ⁉️

I believe the authors are interested in telling us about ✅ clinical improvement in this cohort of patients taking remdesivir, in particular they want to estimate the cumulative incidence of clinical improvement by 28 days.

2/13
They define "clinical improvement" as:

☝️ being discharged alive or
✌️ having a decrease of 2 points or more in a 6-level ordinal scale of oxygen support:

☹️ ECMO
🙁Mechanical ventilation
😕 NIPPV
😐 High-flow oxygen
🙂 Low-flow oxygen
😊 Ambient air

3/13
14 May 20
🧪 Here’s an overview of #COVID19 testing by @LKucirka, @justinlessler, & co on the relationship between the false negative rate (RT-PCR) & when you were tested. This is a *crucial* graph showing how your *pretest* probability of infection is updated by the test result over time
Let’s focus on one of the lines: the dashed line indicating that the pretest probability is 11%

🏡 Context: A recent large study of household contacts estimated the *pretest* probability of infection given someone in your house tested positive is 11.2%
Looking at this graph, if

🏡 someone in your household tests positive
📆 you are tested on the 1st day of exposure
🧪 you test NEGATIVE (phew!)
🦠 your probability of being infected *even though you tested negative* is still 11%

WHY? The test is bad at detecting early results
23 Apr 20
It can be very challenging to assess what information to believe. One simple way to evaluate evidence is via something called "Hill's Criteria", 9️⃣ considerations to help assess whether an observation has a *causal* component

I'll describe them using xkcd comics 🎉

1/12
Sir Austin Bradford Hill, a statistician & epidemiologist, created a list of guidelines for evaluating whether there is evidence of a causal relationship. He determined the following aspects ought to be considered when assessing causality

ncbi.nlm.nih.gov/pmc/articles/P…

2/12
1️⃣ Strength 💪

👉How big is the effect you are seeing?
👉Note: Hill suggests that huge effects can suggest causality, however this does not mean small effects cannot

xkcd.com/539/

3/12
9 Apr 20
I've seen a few papers describing the characteristics of people who tested positive for COVID-19 and this is sometimes being interpreted as describing people with certain characteristic's the *probability of infection*. Let's talk about why that's likely not true 👇🧵

1/22
👉 Usually when thinking about estimating the prevalence of a disease, we use the *sensitivity* and *specificity* of the test to help us
👉 The calculations assume that everyone is equally likely to get tested, and with COVID-19 that is likely not the case

2/22
Let's do some 💭 thought experiments. For these, my goal is to estimate the probability of being infected with 🦠COVID-19 given you have 🧩Disease X

For example,🧩 Disease X could be:
♥️ heart disease
🩸 hypertension
➕ it could also be any subgroup (for example age, etc)

3/22